Chi-square values < 1

Ab initio modelling: DAMMIF, DAMMIN, GASBOR, MONSA
Post Reply
Message
Author
BC_Flügge
Guest
Posts: 2
Joined: 2018.06.15 10:04

Chi-square values < 1

#1 Post by BC_Flügge » 2018.06.15 10:16

Hello,

I try to build ab initio models with DAMMIF/GASBOR and I am always getting Chi-squares values < 1 (usually, around 0.4 - 0.6) no matter what. In my opinion, the raw data looks really good, but I am just a SAXS beginner. I also agree with the generated models as I have a crystal structure to check. I already read that too low Chi-square values are a matter of over-fitting, but I used the data as autognom generated them without changing anything. I will enclose the raw data and images of the fitted curves. As I really want to use this data, I hope for your help. Thank you in advance.

Cheers!
Attachments
HtrASn_RawDataCurve.png
HtrASn_RawDataCurve.png (9.34 KiB) Viewed 131 times
HtrASn_Dammif.png
HtrASn_Dammif.png (11.81 KiB) Viewed 131 times
HtrASn.dat
(22.27 KiB) Downloaded 9 times

User avatar
AL
Administrator
Posts: 642
Joined: 2007.08.03 18:55
Location: EMBL Hamburg, Germany
Contact:

Re: Chi-square values < 1

#2 Post by AL » 2018.07.23 16:08

BC_Flügge wrote:
2018.06.15 10:16
I try to build ab initio models with DAMMIF/GASBOR and I am always getting Chi-squares values < 1 (usually, around 0.4 - 0.6) no matter what.
The experimental errors (3rd column in your HtrASn.dat file) are clearly overestimated - that's why the chi-squared values are too low.

BC_Flügge
Guest
Posts: 2
Joined: 2018.06.15 10:04

Re: Chi-square values < 1

#3 Post by BC_Flügge » 2018.08.03 07:54

Thank you for the reply. And what can I do to avoid overestimation?

User avatar
AL
Administrator
Posts: 642
Joined: 2007.08.03 18:55
Location: EMBL Hamburg, Germany
Contact:

Re: what can I do to avoid overestimation?

#4 Post by AL » 2018.08.03 12:51

Redo the radial averaging. E.g. if you use pyFAI - switch "pixel splitting" to "No splitting".

Post Reply